1189

Towards open-ended evolution in self-replicatingmolecular systemsHerman Duim1 and Sijbren Otto*2

Review Open Access

Address:1Zernike Institute for Advanced Materials, University of Groningen,Nijenborgh 4, 9747 AG Groningen, The Netherlands and 2StratinghInstitute for Chemistry, University of Groningen, Nijenborgh 4, 9747AG Groningen, The Netherlands

Email:Sijbren Otto* - S.Otto@rug.nl

* Corresponding author

Keywords:autocatalysis; open-ended evolution; origin of life; self-replication;synthetic life

Beilstein J. Org. Chem. 2017, 13, 1189–1203.doi:10.3762/bjoc.13.118

Received: 06 December 2016Accepted: 18 May 2017Published: 21 June 2017

This article is part of the Thematic Series "From prebiotic chemistry tomolecular evolution".

Guest Editor: L. Cronin

© 2017 Duim and Otto; licensee Beilstein-Institut.License and terms: see end of document.

AbstractIn this review we discuss systems of self-replicating molecules in the context of the origin of life and the synthesis of de novo life.

One of the important aspects of life is the ability to reproduce and evolve continuously. In this review we consider some of the

prerequisites for obtaining unbounded evolution of self-replicating molecules and describe some recent advances in this field.

While evolution experiments involving self-replicating molecules have shown promising results, true open-ended evolution has not

been realized so far. A full understanding of the requirements for open-ended evolution would provide a better understanding of

how life could have emerged from molecular building blocks and what is needed to create a minimal form of life in the laboratory.

1189

IntroductionMankind has always pondered upon its own existence and has

sought to understand the origin of life. This led us to trace back

our roots, from the great apes to a last universal common

ancestor, a simple cellular lifeform from which all other

present-day organisms have descended. Ultimately this leads us

to one of the great questions in science; how can life emerge

from inanimate matter? And even more interestingly, can we

achieve such a process in the lab and create life from scratch?

There are many different theories surrounding the origin of life

and several attempts have been made to realize the synthesis of

de novo life. All theories involve the presence of molecules that

can create copies of themselves at some stage. It remains

unclear whether such molecules were already important at the

very early stages of the origin of life or whether life started with

large autocatalytic networks [1] and specific molecules that

store genetic information only appeared later. These self-repli-

cating molecules carry hereditary information in the form of

their molecular structure that can be passed on to successive

generations. If mutations occur during the replication process,

genetic information can change from one generation to the next.

Natural selection can act on these variations, favoring those

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1190

varieties that are beneficial for the stability and reproduction of

the replicator. Under the right conditions, such Darwinian type

evolution can eventually lead to diversification and complexifi-

cation of the molecules in the system.

This review aims to provide an insight into the historical back-

ground and recent developments in the field of in vitro evolu-

tion of self-replicating molecules. To do so, we will first cover a

few important principles of Darwinian evolution and will show

how these concepts apply to the case of molecular self-replica-

tion. This is followed by a description of some self-replicating

systems and their properties, starting from the very first report

on self-replication to more elaborate systems. Finally, some

recent experiments concerning in vitro evolution of self-repli-

cating molecules and networks will be discussed. We argue

that, although systems that show intriguing evolutionary capa-

bilities have been devised, there is still a long way to go before

a system that is capable of true undirected or open-ended evolu-

tion has been realized. Worryingly, the phenomenon of open-

ended evolution in itself is currently not well-defined nor under-

stood. If we are to create life in the lab, a thorough knowledge

of this concept and its prerequisites is probably essential.

Review1 Requirements for Darwinian evolutionOne of the most remarkable and key features of life is the fact

that it has a strong tendency (or, at least ability) to diversify and

increase in complexity. Whereas life once must have started out

as a comparatively simple and primitive form, it has diversified

into a vast variety of species ranging from aquatic to airborne

ones. The principles governing this diversification in biological

systems were already described by Darwin in his famous work

On the Origin of Species, but are still not understood in full

detail [2]. It was only in the 1960’s that Spiegelman extended

the scope of Darwinian evolution to chemical systems by

studying the evolution of RNA-complexes [3]. In these experi-

ments RNA was replicated using enzymes “borrowed” from

contemporary biology. The outcome of the selection experi-

ments was the shortening of the RNA sequence, as shorter se-

quences could be replicated faster. It was soon realized that a

better understanding about how evolution acts on the molecular

level would not only provide valuable insights into the origin of

life and the emergence of species, but it could also pave the way

towards the realization of synthetic life.

In biology, Darwinian evolution in a chemical system can be

considered to be the result of an interplay of the three different

processes that are summarized in Figure 1 [4]. These concepts

can, in principle, be extended to what we will consider as

Darwinian evolution in chemical systems. First the parent mole-

cule, or replicator, is replicated to yield a large number of

copies. This can for instance be achieved via an autocatalytic

cycle, as will be discussed below. Mutation involves the emer-

gence of a difference between the parent template and its

copies. The accuracy of the replication process of DNA is gen-

erally safeguarded by sophisticated enzymes, but systems that

lack such machinery are more prone to occasional errors during

replication. Mutations, however, might actually be advanta-

geous for the replicator if the newly formed copies are more

stable, replicate more efficiently or prevail under a change of

environment. If such advantageous mutations arise, competi-

tion between different replicators might occur, leading to a

process of natural selection and survival of the fittest replicator.

We consider replication, mutation and selection to be necessary

and sufficient conditions for Darwinian evolution.

Figure 1: Three processes involved in Darwinian evolution. Speciesmust be replicated to obtain a large population. During the replicationprocess mutations can occur, on which natural or artificial selectioncan then take place.

1.1 ReplicationDuring replication a large number of copies of the replicator is

produced via an autocatalytic process. In animals or other life-

forms it is quite clear that reproduction leads to a transfer of

genetic information from the parent to the offspring. Not only

are the vital structures of the organism transferred but also some

peculiarities like a specific eye color or a hereditary disease is

passed on to the next generation. The transfer of information in

replicating molecules may be less obvious, but if one considers

a polymer with a specific sequence of subunits, it is clear that

some form of genetic information is transferred if the copies

have an identical sequence to the parent molecule.

The survival of a particular molecular structure under a set of

environmental conditions depends on both the rate of replica-

tion and the rate of decomposition of this replicator. If a repli-

cator decomposes at a higher rate than that it is produced, that

particular replicator may become extinct. If, on the other hand,

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1191

the sequence and structure of the molecule is such that the repli-

cation rate exceeds its destruction rate, the replicator is suffi-

ciently adapted to its environment and will persist under the

given conditions.

1.2 MutationThe environmental conditions, to which a set of replicators is

exposed however, may not be in a steady-state. Instead, the

environment may be continuously changing. Consider for

instance changes in temperature, acidity, light intensity and

humidity to which a system will inevitably be subjected. A

replicator that is very well adapted to a certain environment,

might not persist at a later time when the environmental condi-

tions have changed. In fact, if it were not for the presence of

small mutations in the genetic information, a species would be

very fragile. If the number of mutations is only small compared

to the length of the molecule, hereditary information is still

largely preserved. Moreover, the mutants may be better adapted

to the new environmental conditions than their predecessors. If

the mutants indeed have a higher rate of accumulation than their

parents, they will eventually overtake their parents and will

become the new dominant species. In practice a replicator gen-

erally has to exhibit exponential growth in order to dominate

over a weaker replicator [5-7].

Eigen et al. noted, however, that the case is somewhat more

complicated than a single type of mutant replicator overtaking

another replicator. They introduced the concept of quasi-

species, as an analogue to conventional species in biology [8].

A quasi-species consists of a master sequence with a dynamic

distribution of closely related mutants. This concept captures

the fact that for relatively high mutation rates not a single fittest

replicator, but rather a distribution of closely related mutants

survives. The mutants in this distribution around a master se-

quence all replicate at a different rate and are cross-catalytic,

which leads to the production of further mutants. Selection in

these systems thus does not act on the level of individual

mutants, but rather on the entire quasi-species [8-10]. Such

quasi-species behavior was recently reported in an in vitro

evolution experiment with replicating RNA species [11].

There is of course a constraint on the number of mutations that

can occur without losing too much hereditary information from

the parent molecules. In the same work, Eigen showed that

unless mutation rates were significantly diminished, the

increase in the length of the genome would unavoidably lead to

a catastrophic loss of hereditary information. That is, the repli-

cation process of a long molecule requires a much higher

fidelity than that of a smaller molecule. If the rate at which

errors in the replication occur exceeds a certain error threshold,

the genetic information will disintegrate and the species will go

extinct [8,12]. In fact, the reason that viruses are so good in

adapting to different environments and always seem to be one

step ahead of the defense mechanisms of the host is because the

replication process of the viral genome operates very close to

the error threshold, allowing for as many mutations as possible

without the loss of genetic information [12,13].

1.3 SelectionIn biology natural selection operates on the phenotype, i.e., the

observable traits of a species. An individual that is better

adapted to its environment is more likely to survive then one

that is less adapted. This higher survival rate will lead to a

larger amount of offspring for that type of individual, favoring

their presence in the population. The phenomenon of natural

selection can also operate at the molecular level. This requires

experiments to be run under conditions where replication and

replicator destruction occur in parallel. Such conditions were

employed in only a small subset of the work on self-replicating

molecules where the emphasis has mostly been on replication in

the absence of destruction. Which replicators then end up being

selected depends on their rates of replication relative to their

rate of destruction, or, as proposed by Pross, their dynamic

kinetic stability [14]. Selection in the Darwinian sense requires

extinction of the weaker replicators, so that only the fitter ones

remain. There are some detailed kinetic considerations that lead

to specific mechanistic requirements for the replication process.

Szathmary and Lifson showed that in a scenario where differ-

ent replicators compete for common building blocks, extinction

of the weakest replicators occurs only if the kinetic order of the

replicator in the replication process is at least equal to the order

of the replicator in the destruction process [5,6]. As for most

plausible mechanisms the destruction process is first order in

replicator, this implies that the replication process must also be

at the least first order in replicator; i.e., replicators need to be

able to grow exponentially in order to exhibit Darwinian evolu-

tion in the most common scenarios. This consideration has

spurred many efforts to develop exponential replicators, which

are far from trivial to produce (vide infra). But even with expo-

nential replicators, Darwinian evolution does not necessarily

lead to complexification and the spontaneous emergence of new

function, as the Spiegelman experiments made painfully clear

[3].

Yet, in order to obtain a form of life from a molecular system, it

must be able to grow increasingly complex and diverse.

Systems that undergo such undirected diversification may in the

end give rise to ecosystems full of complex organisms or struc-

tures [15]. Note that these organisms then would all be part of

an evolving ecosystem, and it has been argued that a proper and

complete description of life should therefore not only be at the

individual level but also at the level of entire ecosystems [16].

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1192

1.4 Dynamic kinetic stabilityPross has introduced the useful concept of dynamic kinetic

stability for describing the fate of systems in which replication

and selection occur concurrently [14]. The idea is that the

stability of a self-replicator in a system in which replication and

destruction processes occur simultaneously is not determined by

the thermodynamic stability of the replicator, nor by the rate of

formation of the replicator alone, but by the balance between

the rate of formation and the rate of destruction of the repli-

cator. As either replication or destruction (or both) are typically

coupled to other chemical reactions that convert high-energy

reactants to low-energy products, replication in a replication/

destruction regime should normally be chemically fueled. Such

fueling, in principle, allows complexification of the replicator,

without defying the second law of thermodynamics, as the

system as a whole still evolves towards increasing entropy.

With replicator complexification having been made feasible, it

then only depends on evolutionary possibilities and benefits

whether complexification also actually occurs.

In order for the considerations of dynamic kinetic stability to

apply and in order for Darwinian evolution to occur, it is essen-

tial that replicators are subjected to a replication–destruction

regime. Unfortunately, until now, very few systems reported in

the literature are (see below).

1.5 Open-ended evolutionAs mentioned before, the Darwinian triad of replication, selec-

tion, and mutation in itself is not sufficient to drive the

complexification of a chemical system. But what determines

whether a (chemical) system is capable of growing in complexi-

ty or is condemned to remain at a low level of complexity? This

is a question that is not only relevant in evolutionary chemistry,

but also has far-reaching consequences for the development of

artificial life in computer models. As Moreno and Ruiz-Mirazo

point out, in order for a system to fully evolve it should not only

exhibit structural variety, but also some form of functional

variety [17]. In the context of this review, we consider such

function as any property of the replicator that benefits the

dynamic kinetic stability of the system as a whole. A term that

is widely used to describe the emergence of novel functionality

is that of open-ended evolution. Although a clear consensus

about a definition of open-ended evolution is lacking in litera-

ture, we will adopt the definition provided by Taylor here.

Open-endedness means the capability of components in a

system to develop new forms continuously [18]. From this defin-

ition it follows that a self-replicating system should be able to

explore a huge number of possible mutants, otherwise the

system will either get trapped in a stationary optimum situation

or will recycle already explored forms of the replicator [19,20].

Both of these situations cannot lead to the continuous develop-

ment of new forms of replicators and are thus detrimental to the

open-endedness of the system. Another requirement is that the

total structural space available to the system should exceed by

many orders of magnitude the actual structural space that the

system occupies at any one time, or as Maynard-Smith and

Szathmáry put it; the replicators should possess unlimited

heredity [21]. It is also important to note that newly evolved

replicators are not necessarily more advanced or better than the

original replicator. It is the mere development of novelty that is

the vital aspect of open-ended evolution, causing it to be an

undirected process that does not necessarily entail progress

[18].

It is however not that trivial that a replicator can give rise to

such a large number of new forms. As Crutchfield and Schuster

pointed out, the dichotomy of genotype and phenotype is a

powerful mechanism to obtain such a vast number of possible

mutants [22]. Since mutations act on the genotype only and

selection pressure exclusively acts on the phenotype, the two

mechanisms are partially decoupled. If this were not the case

only those mutations that are favored by selection will occur,

strongly decreasing the possible number and randomness of

new forms of the replicator.

It is apparent that open-ended evolution plays an important role

in the emergence of novelty from simple replicators and that

Darwinian evolution alone is an insufficient requirement for

true unbounded evolution in a chemical system. This undirec-

tional evolutionary process is therefore considered to be of

importance in the transition from inanimate matter to life. The

exact principles governing open-ended evolution are however

not yet fully understood and it is not clear what the precise

requirements for a system are in order for it to be capable of

open-endedness [23].

In the following section we will discuss some basic principles

of self-replication, followed by a discussion on recent develop-

ments towards the realization of open-ended evolution in chem-

istry.

2 Replicating systemsThe most instructive and intuitive self-replicating system to

consider is probably that of DNA. A DNA molecule consists of

two strands of nucleotides that are intertwined to form a double

helix. During the replication process of DNA, each of these

strands can act as a template for the formation of a complemen-

tary strand. In this way an exact copy of the original structure of

DNA is formed and the DNA has successfully become repli-

cated. The replication of DNA however is a complex process

mediated by enzymes such as DNA polymerase and topoisom-

erase. To better understand the origin of life and as a possible

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1193

first step in the synthesis of de novo life it would be very inter-

esting indeed to achieve molecular replication without the need

of such enzymes, since enzymes themselves must be products

of an evolutionary process and can thus not explain the emer-

gence of living systems from basic chemical building blocks.

The following section will treat a representative selection of

self-replicating systems, for a more comprehensive overview,

see: Philp and Vidonne [24], Von Kiedrowski and Bag [25] and

Bissette and Fletcher [26].

2.1 Minimal self-replicating systemThe simplest form of a self-replicating system is that in which

the replicator acts as a catalyst for its own formation from a set

of basic building blocks. This fundamental form of a self-repli-

cating system is depicted in Figure 2 and is called a minimal

self-replicating system. An essential requirement for a minimal

replicating system is that molecules A and B are complementa-

ry to template T so that they are able to bind to it via noncova-

lent interactions.

Figure 2 shows three different channels in a minimal repli-

cating system. Building blocks A and B can react via the bimol-

ecular reaction pathway, to form the template molecule T. In

the second pathway – binary complex formation – A and B bind

together reversibly to form a complex [A∙B]. This complex may

undergo a covalent reaction if A and B experience an increased

effective molarity, leading to an inactive template Tinactive

which is folded back onto itself. The third pathway in the

minimal replication system is the autocatalytic cycle. In this

cycle, the building blocks A and B bind reversibly to the com-

plementary recognition sites on the template molecule T. This

arrangement brings molecules A and B in close proximity,

leading to an increased effective molarity and enhanced rate of

bond formation. When A and B ligate to each other, a [T∙T]

complex is formed, which can then dissociate to yield two iden-

tical T molecules. The autocatalytic cycle thus leads to a repli-

cation of the original template molecule. The final, and often

overlooked, pathway was identified by Reinhoudt et al.

following a fierce discussion between Rebek and Menger about

the mechanisms involved in the self-replication in their systems

[27]. In this pathway (not depicted in the diagram) one of the

building blocks, say A, binds to the template molecule. In

certain systems this can lead to the activation of A, such that B

can then react with A directly from solution.

Initially, when there are virtually no template molecules in the

mixture but only building blocks A and B, the bimolecular and

binary complex pathways leading to the formation of T and

Tinactive will be dominant. Clearly, the inactive template cannot

lead to autocatalysis and therefore hinders the self-replication

process. Upon formation of T, the autocatalytic pathway will

Figure 2: Minimal system for self-replication. Building blocks A and Bcan react to form either template T or its inactive counterpart Tinactive.The formation of template T can direct A and B to a configuration inwhich they are in close proximity, accelerating the reaction between Aand B leading to the formation of the [T∙T] complex. Dissociation of thiscomplex completes the replication cycle of the initial template mole-cule.

become increasingly important, in principle allowing for expo-

nential growth of the template. A requirement for effective

autocatalysis, however, is the dissociation of the [T∙T] complex

into two individual template molecules. If this complex does not

dissociate, the newly formed template molecule cannot lead to

further enhancement of the reaction rate, effectively arresting

the autocatalytic cycle. Such product inhibition is an important

limiting factor in many synthetic replicator systems and

prevents them from attaining exponential growth.

2.2 Reciprocal self-replicating systemA more complicated situation arises when the template mole-

cules under consideration are no longer self-complementary, but

instead are complimentary to a second template molecule. The

replication of DNA is a prime example of such a reciprocal self-

replicating system. One strand of the double helix acts as a tem-

plate for the formation of the other complementary strand and

vice versa. Figure 3 shows a schematic representation of a reci-

procal replicating system. It consists of two catalytic cycles

which both lead to the same template duplex [TCD∙TEF].

Instead of only two building blocks the reciprocal system has

four basic building blocks labeled C, D, E and F. Building

blocks C and D, can react to form the template TCD which

catalyzes the formation of the complementary template TEF

from building blocks E and F. Similarly the TEF template can

promote the formation of the TCD template.

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1194

Figure 3: A cross-catalytic replication scheme in which the formationof one template stimulates the formation of a different, complementarytemplate.

2.3 Reaction kinetics and its implicationsWhen considering a mixture containing only building blocks A

and B in the minimal replicator model (Figure 2), the formation

of template molecules T can initially only take place via the

bimolecular reaction pathway. The bimolecular reaction is a rel-

atively slow reaction, since it involves the unassisted formation

of a covalent bond between the two reactants. However, if a

sufficiently large amount of the template molecules is formed,

the autocatalytic cycle will play an increasingly dominant role.

Because the catalytic cycle leads to a doubling of the template

molecules after each run, an exponential increase in the concen-

tration of the reaction product would be expected. Naturally,

this exponential increase cannot continue indefinitely and will

slow down as the concentration of available building blocks

decreases. In summary, this would mean that for an idealized

minimal self-replicating system the concentration of the reac-

tion product T would show an S-like or sigmoidal shape.

A system in which product inhibition occurs, will not show

exponential growth (for exponential growth the kinetic order in

replicator r = 1) but only sub exponential growth. In many

cases r = 1/2 and the system is said to obey the square root law

of autocatalytic systems [28].

Figure 4: The first oligonucleotide capable of template directed self-replication without the need of enzymes. The depicted hexamer tem-plate T is formed from two trimer building blocks and catalyzes its ownformation. Self-reproduction of this molecule was shown to result inparabolic growth of the template concentration [30].

By seeding mixtures with different amounts of preformed tem-

plates T and measuring the initial rate of template formation, a

plot of log(d[T]/dt) versus log[T] can be constructed. From the

slope of this plot the reaction order r of the system can be deter-

mined [28]. It should however be considered that if the uncata-

lyzed bimolecular pathway (r = 0) also contributes to the for-

mation of T, the measured reaction order r reflects a weighted

average of the catalyzed and uncatalyzed pathways and can

therefore have a value smaller than 1, even for cases where the

autocatalytic pathway itself would have a reaction order r = 1.

In such situations computational simulations of the system can

provide additional information on the replication processes that

are involved [29].

2.4 Achieving exponential replicationPioneering work in the field of non-enzymatic self-replication

has been performed by the group of von Kiedrowksi, who was

the first to report on a template-directed self-replicating oligo-

nucleotide (Figure 4) [30]. To achieve template-directed self-

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1195

replication without the aid of enzymes, they used two trinu-

cleotides. Upon activation, these trinucleotides can condense to

form a hexamer template molecule T, depicted in Figure 4,

which catalyzes its own formation. The autocatalytic nature of

the reaction was proven by adding small amounts of preformed

template molecules to the reaction mixture. Kinetic analysis

revealed that the system exhibits parabolic replication (p = 1/2).

Exponential growth in this system is not obtained due to the

high thermodynamic stability of the [T∙T] dimer, leading to

product inhibition. Although the efficiency of the reported auto-

catalytic cycle is rather low, it still was a clear demonstration of

a template-directed self-replicating system and von Kiedrowski

did not fail to recognize the potential of natural selection in

such systems.

Later research focused on overcoming the product inhibition

problem in order to obtain exponential instead of parabolic

growth of the replicators. A successful approach to overcoming

product inhibition involves the immobilization of the template

molecules by fixing them onto a solid support. This approach

was partially inspired by the notion that surfaces of minerals

might have played a major role in catalyzing the formation of

biopolymers [31,32]. Von Kiedrowski et al. were able to

demonstrate exponential growth of oligonucleotides using a

method that they gave the eloquent anagram; SPREAD (Sur-

face-Promoted Replication and Exponential Amplification of

DNA analogues) [33]. In the SPREAD technique, depicted in

Figure 5, an oligonucleotide template strand is immobilized via

an irreversible interaction with a solid support. A complementa-

ry strand is then produced via the template-directed binding of

free nucleotides from the solution. The copied strand is re-

leased from the template and is in turn itself immobilized on a

solid support, thereby preventing product inhibition via the for-

mation of stable template dimers. As von Kiedrowski and

coworkers rightfully notice, this system allows for evolutionary

processes to take place. Moreover, such immobilized systems

are proposed to be even capable of amplification of mutations.

The introduction of mutations can lead to a weaker base pairing

between the template molecule and its copy, thus increasing the

efficiency of the separation of this particular template duplex.

Considering the proposed abundance of amino acids, it is

natural to assume the presence of peptides and oligopeptides

under prebiotic conditions. However, initially only very short

peptides were produced in experiments under such conditions,

raising doubts over their potential role as a precursor of life.

When forming α-helices however, longer polypeptides can be

stabilized by the formation of coiled-coil motifs as in Figure 6.

If every a and d position of each individual helix is occupied by

a hydrophobic amino acid, the helices can intertwine and bury

their hydrophobic side groups into each other. This hydro-

Figure 5: Replication involving the SPREAD technique which preventsproduct inhibition. (1) A template molecule is immobilized on a solidsupport and (2, 3) a complementary copy is produced by template-directed replication. Finally the copied strand is (4) released from thetemplate and is in turn immobilized [33].

Figure 6: Figure showing (a) a coiled coil motif due to hydrophobicinteractions between hydrophobic amino acids in the individual helices.(b) Helical-wheel diagram showing how the hydrophobic amino acidssituated on the a and d sites can interact with each other to form thecoiled coil [34].

phobic interaction that drives the formation of coiled-coil motifs

can be further enhanced by electrostatic interactions between

amino acids residing on the c and g positions of the α-helices.

Ghadiri et al. showed that such coiled-coil peptides are capable

of self-replication [35]. As depicted in Figure 7, helical

polypeptides can act as a template for shorter peptide fragments

by means of molecular recognition. The peptide building blocks

again are ligated, resulting in the formation of a template duplex

with a coiled-coil motif. When separated from the original tem-

plate, a copy of the template is obtained. Initially these repli-

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1196

Figure 7: Self-replication of a helical peptide. Molecular recognitionleads to the formation of a stable coiled coil structure from smallerpeptide fragments. Depending on the length and stability of the coiledcoil structure, the template-duplex dissociates in the original templateand a copy [37].

cating systems were reported to show only parabolic growth,

because of the very high stability of the coiled-coil structure.

This problem was later addressed by Issac and coworkers by

reducing the length of the template molecule, which led to a de-

creased stability of the template duplex. Using this approach

they obtained near exponential growth of the template concen-

tration of p = 0,91 [36,37]. The above examples all illustrate

that, while not trivial, it is indeed possible to obtain self-repli-

cating behavior in the absence of enzymes. While this marks a

significant contribution to our understanding of the early stages

of the transition from chemistry to biology, it does not directly

explain the emergence of the RNA and DNA dominated world

as we know it, which would probably have required open-ended

evolution.

3 Evolutionary dynamics of replicators3.1 Enzyme mediated replicationIconic early experiments aiming to achieve Darwinian type

evolution in a chemical system were performed by Spiegelman

et al. in 1967 [3]. RNA replicase and a small input of genomic

RNA were successfully isolated from the bacteriophage Qβ.

The RNA molecules in this system are replicated by an RNA

replicase enzyme. By successive rounds of amplification and

selection, selection pressure was introduced to the system by

favoring fast reproducing entities of the genomic RNA. Since

shorter sequences are being replicated at a higher rate than

longer sequences, shortened mutants are favored over longer se-

quences. This eventually led to a strong decrease in the genome

size of the RNA molecules. However, this result is not as trivial

as it may seem at first sight, since it is of vital importance that

the mutant species do not lose their ability to be replicated, indi-

cating that only specific parts of the genome that are not needed

for recognition by the polymerase were deleted. Although the

RNA molecules involved are not self-replicating but are repli-

cated by the RNA replicase, the study still marks a starting

point in the field of in vitro evolution. Later, Braun et al.

managed to apply a selection pressure that favors the replica-

tion of long DNA sequences over short strands by creating heat

gradients in pores that act as a thermal trap [38]. The thermal

traps selectively retain longer DNA sequences, thereby effec-

tively overcoming the inherent advantage of the replication of

short sequences.

3.2 Dynamics of self-replicatorsAshkenasy recently reported a peptide based synthetic autocat-

alytic network that shows two significantly distinct steady states

depending on the history of the system [39]. Depending on the

initial concentration of replicator molecules provided to the

system, the system will reach either a low or a high steady state

replicator concentration. Switching between these two states

can be achieved by applying external stimuli in the form of heat

or the addition of decomposing agents. The switchable behav-

ior and memory of such a self-replicating system constitute an

exciting step, moving systems of self-replicating molecules

away from equilibrium, with potential impact on evolutionary

behavior [40].

Another interesting dynamic emergent property of self-repli-

cating systems was demonstrated by Philp and coworkers [41].

They showed how self-replicating molecules can create a reac-

tion-diffusion front when seeded to a homogeneous mixture of

building blocks. Dynamically evolving out-of-equilibrium envi-

ronments like these could enable interesting behavior of replica-

tors that is not achievable in homogenous reaction mixtures. It

will be very exciting to observe the evolutionary behavior of

mixtures of replicators in such spatially resolved environments.

3.3 RNA self-replicationOwing to the importance of RNA in viral species and in the

origin of life, evolution experiments are most often performed

using RNA molecules or closely related derivatives. In fact, it

has become possible to perform natural selection on oligo-

nucleotides by iterative amplification and selection processes

using a technique called systematic evolution of ligands by

exponential enrichment, or SELEX. In SELEX, a library of

DNA and RNA sequences is exposed to a certain target. In

multiple selection rounds the binding species are selected and

amplified, while the non-binding DNA and RNA molecules are

disposed of. In this way molecules are evolved based on their

ability to bind to a specific target [42,43].

However, these in vitro evolution experiments all exploit RNA-

based enzymes (ribozymes) or proteins in their replication

process to obtain exponential growth and are consequently not

self-replicating. Efforts have been made to obtain in vitro evolu-

tion of RNA in the absence of any enzymes. Unfortunately, the

demonstration of multiple cycles of non-enzymatic RNA repli-

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1197

Figure 8: (a) Cross-catalyzed replication of template molecules E and E’ from their building blocks A(’) and B(’). (b) Secondary structure of the tem-plate duplex. The curved arrow denotes the site of ligation of the building blocks, the dashed boxes include the sequences to be mutated and the solidlines indicate the sites of the G∙U base pairs inducing the wobble that enhances catalytic activity. (c) The altered sequences of the 12 different tem-plate molecules, the E’ molecules have complementary sequences in the base pairing part (horizontal) and identical sequences in the catalytic part(vertical). Dark circles denote the differences relative to E1 [45].

cation in a test tube is troubled by the fact that the RNA duplex

that is formed upon replication is quite stable and can have

dissociation temperatures as high as 90 °C [44]. Without an en-

zyme that separates the newly created strands, this stability

would lead to product inhibition, halting the self-replication

process.

3.4 Cross-catalyzing RNA replicatorsJoyce and Lincoln showed, however, that a system of two RNA

enzymes can catalyze each other’s synthesis from a mixture of

four different building blocks via template-directed reciprocal

replication [45]. The RNA ligase molecule E can bind two

oligonucleotide building blocks A’ and B’ and promote their

ligation to form the ligase E’. The newly formed ligase E’ can

then in turn promote the formation of E, as depicted in

Figure 8a. But this cross-catalytic reaction typically occurs at a

very slow rate. In order to enhance this rate, enzymatic in vitro

evolution of the RNA molecules was performed in order to

obtain a set of fast replicating species.

It was found from in vitro evolution experiments that the intro-

duction of G∙U base pairs close to the site of ligation leads to

enhanced cross-catalytic activity. Figure 8b shows the se-

quence and secondary structure of the A∙B∙E’ complex. The site

of ligation is indicated by the curved arrow and the G∙U pairs

that are depicted in a solid box induce a wobble in the sequence

that results in the enhanced catalytic activity. When this wobble

is installed in both enzymes of the cross-catalytic set, exponen-

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1198

tial growth of the system can be achieved over multiple cycles.

With an exponential replicator in hand, Darwinian evolution in

a cross-catalytic system lies within reach. To study this, Joyce

and his team prepared 12 pairs of cross-catalytic enzymes and

their corresponding building blocks, that have alteration in parts

of the sequence denoted by the dashed line in Figure 8b. The

different pairs are denoted as E1 to E12 and are shown in

Figure 8c. It is important to note that mutations between en-

zymes are such that the stability of the ligase duplex due to base

pairing is not altered, but only the catalytic activity and replica-

tion rate are affected. All these enzymes were shown to cross-

replicate, with the E1 pair showing the highest rate of replica-

tion.

A serial transfer experiment was performed on a mixture that

contained the 12 different enzyme pairs and their correspond-

ing 48 building blocks (A1, A1’, B1, and B1’ for pair E1, for

example). In such a serial transfer experiment a small percent-

age, in this case 5%, is transferred to a new reaction mixture

after a replication round took place. This effectively eliminates

the slow replicators that are only present in small quantities in

the mixture so that they tend to go extinct. The transferred repli-

cators, however, are presented with a fresh batch of building

blocks and can continue to replicate. By doing this for multiple

rounds, large amplification factors can be achieved. For this ex-

periment it is important to realize that A1 does not necessarily

have to be ligated to B1, but that it can ligate to any of the other

B-type building blocks, although they may be mismatched to

the template. This freedom of recombination leads to 132

possible combinations of building blocks. After 20 successive

transfers a 1025-fold amplification was reached. A sample of

100 of these clones contained only 7 non-recombinant clones,

whereas the rest were all ligated to building blocks that were

not their original partners. Figure 9 shows the distribution of

different E (dark columns) and E’ (light columns) enzymes in

the final sample. This result shows how fitter replicators can

come to dominate the population after several rounds of ampli-

fication. Fitness of the molecules depends in this case on their

ability to perform cross-catalytic replication with other mole-

cules.

This study by Joyce et al. demonstrates how selection pressure

can lead to certain replicators dominating a population in a

cross-catalytic replication process. However, the environmental

conditions in this experiment are static and the system lacks

open-endedness because the number of building blocks that is

provided to the system restricts the total diversity of the

newly formed species, in this case 12 × 12 different possible

replicators. This will cause the system to reach a steady state in

which no novel forms of the replicator can be explored

anymore.

Figure 9: Distribution of the species present in the reaction mixtureafter 20 serial transfers. E and E’ molecules are represented by thedark and light shaded bars, respectively. Note how certain specieshave come to dominate the population, particularly A5B3. Moreover,only 7 molecules were found to be paired to their original partner (cor-responding to the shaded diagonal) [45].

3.5 Cooperative catalytic systemThe concept of such a cross-replicating system can be readily

extended to higher order systems, involving three, four or even

more components. Eventually, one could envision an entire

network of cross-replicating molecules. Lehman et al. showed

that a mixture of relatively short RNA segments can self-

assemble to form self-replicating ribozymes [46]. These

ribozymes in turn gave rise to spontaneous formation of cooper-

ative networks that were shown to grow faster than the autocat-

alytic replication rate of the individual ribozymes. Moreover,

cooperative systems are generally more stable towards para-

sites then autocatalytic self-replicators and are, in principle, able

to gain in complexity [46,47].

In the study a ribozyme of around 200 nucleotides called

Azoarcus was used. This ribozyme is made from four different

RNA strands (W, X, Y and Z) that can self-assemble cova-

lently in an autocatalytic manner, as depicted in Figure 10a. The

effectiveness of this self-replication process depends on the

ability of the internal guide strand (IGS) to recognize its target.

To form a cooperative set, the Azoarcus ribozyme was frag-

mented in two different pieces in three different ways, creating

three different pairs I1, I2 and I3 which are shown encircled in

Figure 10b. Furthermore, the target and IGS sequences were

altered such that autocatalytic self-replication is minimized. The

sequence was, however, chosen such that the IGS of one pair is

matched to the target sites of the next pair. In this way one

ribozyme, say E1, can catalyze the formation of the next

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1199

Figure 10: (a) Secondary structure of the Azoarcus ribozyme consisting of four different strands of RNA, W, X, Y, and Z. Self-replication is mediatedby recognition of the target sites by the IGS (grey boxes), leading to ligation of the strands. The dashed line indicates the catalytic core of the result-ing ribozyme. (b) Cooperative replicating system. The formation of the covalent ribozyme E3 from the non-covalent I3 complex is catalyzed by E2ribozyme. The formation of this E2 is in turn catalyzed by E1, which is catalyzed by E3, resulting in a cyclic dependence. Numbers above the arrowsdenote the advantage of cooperativity [46].

ribozyme, E2, from its non-covalently bound building blocks

I2. This ribozyme can in turn catalyze the formation of E3 from

its building block and finally, to close the cycle, E3 can cata-

lyze the formation of the E1 ribozyme. This cooperative system

is depicted in Figure 10b and it was observed that a mixture

containing all three pairs resulted in a much higher yield of full-

length RNA (a factor 125) than obtained from the sum of the

isolated pairs, proving that the system replicates in a coopera-

tive manner.

Interestingly, it was shown that in isolation the autocatalytic

replicators (with the IGS programmed to recognize itself) repli-

cated faster than the cross-catalytic system, whereas in a mix-

ture of all different components the cooperative network grows

faster than the selfishly replicating molecules. However, this

result was obtained using deliberately designed pairs with spe-

cific targets. Behavior becomes a lot more fascinating when one

of the nucleotides of the IGS (M) and target sites (N) is random-

ized, creating a mixture of 48 matched and unmatched pairs in

total, as schematically depicted in Figure 11a. After incubation

of all six sets of Figure 11a for several hours, all of these 48

possible sequences were indeed found in the mixture. Initially

the replication is dominated by autocatalytic cycles in which N

and M are complimentary. This initial rise of the autocatalytic

replicators is depicted in Figure 11b by the dashed line with

crosses, the contribution of the two-membered cycles is

depicted by the dashed lines with dots (depicted value ×10). At

later times a transition to the more complex three-membered

cycles was observed as witnessed by the rise of the solid line

(×10.000). After 8 hours, it was observed that replication occurs

increasingly via cooperative cycles and that all genotypes con-

tribute increasingly with time. This result shows how an

initially autocatalytic cycle can give rise to increasingly com-

plex systems of cooperative replication over time. Interestingly,

the overall replication efficiency of the randomized multi-com-

ponent network exceeded that of the engineered 3-component

network in Figure 10b.

To better mimic prebiotic conditions in which iterations over

multiple generations would have occurred, a serial transfer ex-

periment using the same set of replicators was also performed.

In this experiment an aliquot of the reaction mixture is trans-

ferred to a new flask with building blocks every hour, so that

the more stable and fast replicating molecules and networks are

favored. Again a transition from autocatalytic cycles to more

complex systems was observed.

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1200

Figure 11: (a) The different combinations of IGS strands, tags andbreak junction give rise to a total of 48 different pairs. (b) Graphshowing the frequency of autocatalytic replicators (dashed, crosses),the two-membered cycles (dashed, dots; ×10) and the three-mem-bered cycles (solid; ×10.000) over time. Note the emergence of themore complex three-membered cycles at later times [46].

Such cooperative systems are capable of complexification and

natural selection and can therefore be of importance in bridging

the gap between replication of simple short RNA molecules

from nucleotide building blocks and the formation of more

complex ribozymes. The observed cooperative behavior relies

on recognition strands and tags, so that it will only play a role

for the assembly of intermediate-sized oligonucleotides. Small

oligonucleotides would likely still replicate more efficiently via

auto or cross catalytic cycles. At a certain length scale the for-

mation of cooperative systems becomes favorable and these

mechanisms might take over the replication process, allowing

for complexification and diversification of the system. Howev-

er, since the replication of each member is dependent on one or

more other members of the system, the members should all be

in close proximity to each other in order to obtain a stable

system. This requires high concentrations of the reaction mix-

ture, which is of course readily achieved in the laboratory but is

probably less likely under prebiotic conditions. In order to

increase the concentration of replicators locally, a specialized

compartmentalization should act in concert with the coopera-

tive replication system. How such compartmentalization might

occur is another topic entirely and beyond the scope of this

review, but it is proposed that compartmentalization can actu-

ally aid in the evolution of replicating molecules [48-50].

3.6 Diversification of self-replicatorsOther types of molecules than RNA that are capable of self-

replication and information storage are showing interesting

results in the study of open-ended evolution and the synthesis of

life as well [51]. Recently, we have demonstrated a self-repli-

cating system involving peptides capable of diversification

using a systems chemistry approach [52]. Following the

discovery of an exponentially growing self-replicating system

[53], we used two building blocks, 1 and 2, to form a dynamic

combinational library (DCL) of self-replicating molecules.

These building blocks consist of an aromatic core that is func-

tionalized with two thiol groups and a peptide chain

(Figure 12a). Building block 1 and 2 are very closely related to

each other and differ only in a single amino acid of the peptide

chain. These peptide building blocks can then be oxidized to

form macrocycles of different sizes as depicted in Figure 12b.

The design of the peptide chains is such that self-assembly of

the chains into parallel β-sheets is promoted, which in turn leads

to the formation of stacks of macrocycles as shown in

Figure 12c. Growth of these stacks occurs exclusively via the

ends of the fibers and it is therefore not surprising that the reac-

tion rate is strongly dependent on the amount of fibers present

in the mixture. As soon as a fiber reaches a critical length it can

fragment when mechanically agitated. When fragmentation

occurs, the number of available fiber ends is doubled, leading to

an exponential self-replication.

In previous work it was already shown that the less hydro-

phobic building block 2 tends to form larger octameric macro-

cycles than the more hydrophobic building block 1 which forms

hexamers [54]. This is reasonable, since a weaker hydrophobic

interaction provided by 2 would need more individual interac-

tions in order to achieve the same stability as a more hydro-

phobic counterpart 1.

By using a mixture containing two different building blocks

instead of one, the replicators can potentially undergo mutation

by incorporating a different building block into their structures.

A mixture with equal concentrations of both building blocks

was prepared and monitored over the course of 35 days.

Initially a complex mixture of four different trimers and five

different tetramers was observed. After some days, however, a

set of hexamers which was enriched in building block 1 arose in

the mixture (set I) as shown by the red line in Figure 13. As the

emergence of set I depletes the mixture from building block 1

the environmental conditions are essentially changed up to the

point where a second set of hexamers arises which is rich in

building block 2.

It was shown that set I is the ancestor of set II. When macro-

cycles that are rich in building block 2 are exposed at the fiber

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1201

Figure 12: Figure depicting (a) building blocks consisting of a peptide attached to an aromatic ring. Building blocks 1 and 2 differ only in the nature ofthe penultimate amino acid. (b) The building blocks can form macrocycles of different sizes upon oxidation, which can exchange building blocks witheach other. (c) Schematic representation showing how building blocks oxidize to form macrocycles that in turn form stacks due to β-sheet formation.Stacks grow from their ends and fragment upon agitation, leading to more fiber ends and faster growth [52].

ends of set I, they act as a template for the formation of

members of set II. Indeed, significant amounts of set II

members only form when a seed of set I is present that contains

2-enriched members. Set I is therefore able to transfer informa-

tion about its macrocyclic size to set II. This process bears a

crude resemblance to how species originate in biology.

ConclusionSelf-replicating molecules have been remarkably hard to

develop and after 30 years of research there are still only a

handful of efficient self-replicators. Achieving Darwinian

evolution with these systems has proven even more challenging.

The evolutionary potential of many self-replicating molecules is

limited due to the fact that it is difficult to achieve exponential

growth of the replicator. Factors limiting the efficiency of the

self-replication process are the presence of non-autocatalytic

pathways and product inhibition. Methods aiming to minimize

the effect of product inhibition, like the SPREAD technique and

the destabilization of template-duplexes, have successfully been

developed to allow for exponential growth of some simple

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1202

Figure 13: Plot showing the relative concentrations of set I (red), set II(blue) and the link between them; (1)3(2)3 (purple). Note how at firstonly set I is present in the mixture, while at some point this set givesrise to the descendant set II. [52]

replicators. Also mechanical forces may be utilized to break up

larger assemblies of self-replicating molecules and liberate the

assembly edges or fiber ends that promote replication.

The most impressive progress with respect to Darwinian evolu-

tion has been achieved with RNA-based cross-replicators. In

serial transfer experiments changes in replicator populations

were observed that were not immediately predictable and that

favored the most efficient replicators or networks of cooper-

ating replicators. What these systems have not (yet) shown is

the emergence of new functions that contribute to the dynamic

kinetic stability of the replicators.

The true challenge of any in vitro evolution experiment lies in

the realization of a system that has the capability to undergo

open-ended evolution. Such systems can diversify and increase

in complexity and invent new functions indefinitely. Until now,

chemical systems that show evolutionary behavior have

involved relatively simple replicators that only had access to a

very limited structural space of possible mutations. This rapidly

causes the system to be incapable of exploring new structures

and the development of novelty will stagnate. An additional

limitation of simple replicators is the strong relation between

their genotype and phenotype. This lack of dichotomy causes

the mechanisms of mutation and natural selection to couple to

each another, hampering the evolvability of the systems. It is far

from trivial to design a system that is simple enough to be

capable of exponential replication and has a large structural

space of mutations at the same time. Yet a push in this direc-

tion is probably needed, expanding the structural space avail-

able for existing replicators to explore, enabling them to

discover new functions, one of which might eventually be the

decoupling between genotype and phenotype, which would

allow the system to explore a dramatically larger structural and

functional space.

Besides these issues concerning the design of replicators, it is

still not studied in detail how the environment of the replicators

can interact with the evolutionary process. Can environmental

conditions like acidity or temperature, for instance, be an incen-

tive towards the development of novel functionalities in the

replicators? And how is the notion of death introduced in an ex-

periment in which the researcher does not actively intervene

with the system through, for example, serial dilution? In any

true open-ended system replicators interact with the environ-

ment on their own account and are not steered by the experi-

menter to a significant extent.

Thus, the challenge is now to design systems of self- or cross-

replicating molecules that can access and evolve into a vast

structural and functional space and facilitate, by appropriate

design of building blocks and experimental conditions, the

invention of new functions and thereby achieve open-ended

evolution.

AcknowledgementsH. D. is grateful for the scholarship granted by the Zernike

Institute for Advanced Materials. We also acknowledge the

ERC, the NWO and the Dutch Ministry of Education, Culture

and Science (Gravitation Program 024.001.035) and COST

CM1304.

References1. Kauffmann, S. A. At Home in the Universe; Oxford University Press:

New York, USA, 1995.2. Darwin, C. On the Origin of Species; D. Appleton and CO.: New York,

USA, 1871.3. Mills, D. R.; Peterson, R. L.; Spiegelman, S.

Proc. Natl. Acad. Sci. U. S. A. 1967, 58, 217–224.doi:10.1073/pnas.58.1.217

4. Joyce, G. F. Angew. Chem., Int. Ed. 2007, 46, 6420–6436.doi:10.1002/anie.200701369

5. Szathmary, E.; Gladkih, I. J. Theor. Biol. 1989, 138, 55–58.doi:10.1016/S0022-5193(89)80177-8

6. Lifson, S.; Lifson, H. J. Theor. Biol. 2001, 212, 107–109.doi:10.1006/jtbi.2001.2361

7. Joyce, G. F. Annu. Rev. Biochem. 2004, 73, 791–836.doi:10.1146/annurev.biochem.73.011303.073717

8. Eigen, M.; McCaskill, J.; Schuster, P. J. Phys. Chem. 1988, 92,6881–6891. doi:10.1021/j100335a010

9. Ruiz-Mirazo, K.; Briones, C.; De la Escosura, A. Chem. Rev. 2014,114, 285–366. doi:10.1021/cr2004844

10. Eigen, M. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 13374–13376.doi:10.1073/pnas.212514799

11. Arenas, C. D.; Lehman, N. BMC Evol. Biol. 2010, 10, 80.doi:10.1186/1471-2148-10-80

Beilstein J. Org. Chem. 2017, 13, 1189–1203.

1203

12. Biebricher, C. K.; Eigen, M. Virus Res. 2005, 107, 117–127.doi:10.1016/j.virusres.2004.11.002

13. Crotty, S.; Cameron, C. E.; Andino, R. Proc. Natl. Acad. Sci. U. S. A.2001, 98, 6895–6900. doi:10.1073/pnas.111085598

14. Pascal, R.; Pross, A. Synlett 2017, 30–35.doi:10.1055/s-0036-1589403

15. Ruiz-Mirazo, K.; Pereto, J.; Moreno, A. Origins Life Evol. Biospheres2004, 34, 323–346. doi:10.1023/B:ORIG.0000016440.53346.dc

16. Taylor, T. European conference on artificial life 2015, York, UK, July,2015.

17. Moreno, A.; Ruiz-Mirazo, K. Biol. Philos. 2009, 24, 585–605.doi:10.1007/s10539-009-9178-6

18. Taylor, T. J. From Artificial Evolution to Artificial Life. Ph.D. Thesis,University of Edinburgh, U.K., 1999.

19. Von Neumann, J.; Burks, A. W. IEEE Trans. Neural Networks 1966, 5,3–14.

20. Ruiz-Mirazo, K.; Umerez, J.; Moreno, A. Biol. Philos. 2008, 23, 67–85.doi:10.1007/s10539-007-9076-8

21. Szathmáry, E.; Maynard Smith, J. Nature 1995, 374, 227–232.doi:10.1038/374227a0

22. Crutchfield J.P.; Schuster, P. Genotype and phenotype. In EvolutionaryDynamics: Exploring the Interplay of Selection, Accident, Neutrality,and Function, Oxford University Press: Oxford, U.K., 2003; pp 164-169.

23. Taylor, T.; Bedau, M.; Channon, A.; Ackley, D.; Banzhaf, W.;Beslon, G.; Dolson, E.; Froes, T.; Hickinbotham, S.; Ikegami, T.;McMullin, B.; Packard, N.; Rasmussen, S.; Virgo, N.; Agmon, E.;Clarck, E.; McGregor, S.; Ofria, C.; Ropella, G.; Spector, L.;O. Stanley, K.; Stanton, A.; Timperley, C.; Vostinar, A.; Wiser, M.Artif. Life 2016, 22, 408–423. doi:10.1162/ARTL_a_00210

24. Vidonne, A.; Philp, D. Eur. J. Org. Chem. 2009, 5, 593–610.doi:10.1002/ejoc.200800827

25. Bag, B. J.; Von Kiedrowski, G. Pure Appl. Chem. 2009, 68, 2145–2152.doi:10.1351/pac199668112145

26. Bissette, A. J.; Fletcher, S. L. Angew. Chem., Int. Ed. 2013, 52,12800–12826. doi:10.1002/anie.201303822

27. Reinhoudt, D. N.; Rudkevich, D. M.; De Jong, F. J. Am. Chem. Soc.1996, 118, 6880–6889. doi:10.1021/ja960324g

28. Von Kiedrowski, G. Bioorg. Chem. Front. 1993, 3, 113–146.29. Coulomb-Delsuc, M.; Mattia, E.; Sadownik, J. W.; Otto, S.

Nat. Commun. 2015, 6, 7427. doi:10.1038/ncomms842730. Von Kiedrowski, G. Angew. Chem., Int. Ed. Engl. 1986, 25, 932–935.

doi:10.1002/anie.19860932231. Ferris, J. P.; Hill, A. R.; Liu, R.; Orgel, L. E. Nature 1996, 381, 59–61.

doi:10.1038/381059a032. Ferris, J. P.; Ertem, G. Science 1992, 257, 1387–1389.

doi:10.1126/science.152933833. Luther, A.; Brandsch, R.; Von Kiedrowski, G. Nature 1998, 396,

245–248. doi:10.1038/2434334. Armstrong, C. T.; Boyle, A. L.; Bromley, E. H. C.; Mahmoud, Z. N.;

Smith, L.; Thomson, A. R.; Woolfson, D. N. Faraday Discuss. 2009,143, 359–372. doi:10.1039/B915411F

35. Lee, D. H.; Granja, J. R.; Martinez, J. A.; Severin, K.; Ghadiri, M. R.Nature 1996, 382, 525–528. doi:10.1038/382525a0

36. Issac, R.; Chmielewski, J. J. Am. Chem. Soc. 2002, 124, 6808–6809.doi:10.1021/ja026024i

37. Issac, R.; Ham, Y. W.; Chmielewski, J. Curr. Opin. Struct. Biol. 2001,11, 458–463. doi:10.1016/S0959-440X(00)00233-5

38. Kreysing, M.; Keil, L.; Lanzmich, S.; Braun, D. Nat. Chem. 2015, 7,203–208. doi:10.1038/nchem.2155

39. Mukherjee, R.; Cohen-Luria, R.; Wagner, N.; Ashkenasy, G.Angew. Chem., Int. Ed. 2015, 54, 12452–12456.doi:10.1002/anie.201503898

40. Decker, P. Nature (London) 1973, 241, 72–74.doi:10.1038/newbio241072a0

41. Bottero, I.; Huck, J.; Kosikova, T.; Philp, D. J. Am. Chem. Soc. 2016,138, 6723–6726. doi:10.1021/jacs.6b03372

42. Keefe, A. D.; Pai, S.; Ellington, A. Nat. Rev. Drug Discovery 2010, 9,537–550. doi:10.1038/nrd3141

43. Mahlknecht, G.; Maron, R.; Mancini, M.; Schechter, B.; Sela, M.;Yarden, Y. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 8170–8175.doi:10.1073/pnas.1302594110

44. Hernandez, A. R.; Piccirilli, J. A. Nat. Chem. 2013, 5, 360–362.doi:10.1038/nchem.1636

45. Lincoln, T. A.; Joyce, G. F. Science 2009, 323, 1229–1232.doi:10.1126/science.1167856

46. Vaidya, N.; Manapat, M. L.; Chen, I. A.; Xulvi-Brunet, R.; Hayden, E. J.;Lehman, N. Nature 2012, 491, 72–77. doi:10.1038/nature11549

47. Higgs, P. G.; Lehman, N. Nat. Rev. Genet. 2015, 16, 7–17.doi:10.1038/nrg3841

48. Ghadessy, F. J.; Ong, J. L.; Holliger, P. Proc. Natl. Acad. Sci. U. S. A.2001, 98, 4552–4557. doi:10.1073/pnas.071052198

49. Szostak, J. W.; Bartel, D. P.; Luisi, P. L. Nature 2001, 409, 387–390.doi:10.1038/35053176

50. Matsumura, S.; Kun, A.; Ryckelynck, M.; Coldren, F.; Szilágyi, A.;Jossinet, F.; Rick, C.; Nghe, P.; Szathmáry, E.; Griffiths, A. D. Science2016, 354, 1293–1296. doi:10.1126/science.aag1582

51. Pinheiro, V. B.; Taylor, A. I.; Cozens, C.; Abramov, M.; Renders, M.;Zhang, S.; Chaput, J. C.; Wengel, J.; Peak-Chew, S. Y.;McLaughlin, S. H.; Herdewijn, P.; Holliger, P. Science 2012, 336,341–344. doi:10.1126/science.1217622

52. Sadownik, J. W.; Mattia, E.; Nowak, P.; Otto, S. Nat. Chem. 2016, 8,264–269. doi:10.1038/nchem.2419

53. Carnall, J. M. A.; Waudby, C. A.; Belenguer, A. M.; Stuart, M. C. A.;Peyralans, J. J. P.; Otto, S. Science 2010, 327, 1502–1506.doi:10.1126/science.1182767

54. Malakoutikhah, M.; Peyralans, J. J. P.; Colomb-Delsuc, M.;Fanlo-Virgós, H.; Stuart, M. C. A.; Otto, S. J. Am. Chem. Soc. 2013,135, 18406–18417. doi:10.1021/ja4067805

License and TermsThis is an Open Access article under the terms of the

Creative Commons Attribution License

(http://creativecommons.org/licenses/by/4.0), which

permits unrestricted use, distribution, and reproduction in

any medium, provided the original work is properly cited.

The license is subject to the Beilstein Journal of Organic

Chemistry terms and conditions:

(http://www.beilstein-journals.org/bjoc)

The definitive version of this article is the electronic one

which can be found at:

doi:10.3762/bjoc.13.118